Methanosarcinaceae and Acetate-Oxidizing Pathways Dominate in High-Rate Thermophilic Anaerobic Digestion of Waste-Activated Sludge Downloaded from Dang P. Ho, Paul D. Jensen, Damien J. Batstone Advanced Water Management Centre, The University of Queensland, St Lucia, Queensland, Australia

This study investigated the process of high-rate, high-temperature methanogenesis to enable very-high-volume loading during anaerobic digestion of waste-activated sludge. Reducing the hydraulic retention time (HRT) from 15 to 20 days in mesophilic digestion down to 3 days was achievable at a thermophilic temperature (55°C) with stable digester performance and methano- genic activity. A volatile solids (VS) destruction efficiency of 33 to 35% was achieved on waste-activated sludge, comparable to

that obtained via mesophilic processes with low organic acid levels (<200 mg/liter chemical oxygen demand [COD]). Methane http://aem.asm.org/ yield (VS basis) was 150 to 180 liters of CH4/kg of VSadded. According to 16S rRNA pyrotag sequencing and fluorescence in situ hybridization (FISH), the methanogenic community was dominated by members of the , which have a high level of metabolic capability, including acetoclastic and hydrogenotrophic methanogenesis. Loss of function at an HRT of 2 days was accompanied by a loss of the , according to pyrotag sequencing. The two acetate conversion pathways, namely, acetoclastic methanogenesis and syntrophic acetate oxidation, were quantified by stable carbon isotope ratio mass spectrometry. The results showed that the majority of methane was generated by nonacetoclastic pathways, both in the reactors and in off-line batch tests, confirming that syntrophic acetate oxidation is a key pathway at elevated temperatures. The proportion of methane due to acetate cleavage increased later in the batch, and it is likely that stable oxidation in the continuous reactor was maintained by application of the consistently low retention time. on November 12, 2015 by University of Queensland Library

naerobic digestion (AD) is a biological process which has methanogenic activities at a short HRT of 2 to 4 days in the ther- Abeen widely used for waste-activated sludge and primary mophilic pretreatment stage of a two-stage temperature-phased sludge stabilization. It offers substantial advantages over alterna- anaerobic digestion (TPAD) (11) when the system is fed activated tive stabilization techniques, including low energy requirements, sludge. good-quality biosolids product, energy recovery in the form of The main risk with a short HRT is washout of methanogens as methane (CH4) gas, and the possibility of nutrient recovery (1, 2). well as other functional groups, causing a build-up of organic Conventional anaerobic digesters are operated at a mesophilic acids and a decreased pH (12). A key concern regarding elevated temperature (35°C) with long hydraulic retention times (HRTs) temperatures is the decrease in pKa of the ammonium/ammonia ϩ of 15 to 20 days on waste-activated sludge (WAS), which is excess (NH4 /NH3) acid-base pair, which means that the NH3 concen- bacterial material and residual organics from combined aerobic tration will be 2 to 3 times higher for a given pH, potentially treatment of wastewater, and more than 20 days on primary causing ammonia inhibition (1 to 2 mM) (2). sludge (PS), which is the settled fraction from raw wastewater (1, The microbial composition and distribution in anaerobic di- 3, 4). Long treatment times are a key disadvantage of anaerobic gesters have been the focus of recent studies using high-through- digestion, requiring large tank volumes and, consequently, high put molecular techniques. The bacterial and archaeal community capital cost, as well as increased mixing and maintenance costs. structures are highly diverse and affected by substrate composi- Therefore, reducing digester volume would strongly enhance the tion (6, 12–14), HRT (12), and operating temperature (6, 15). competitiveness of anaerobic technologies. Two methanogenic pathways from acetate have been reported, Thermophilic anaerobic digestion (55°C to 70°C) is an alter- including (i) direct acetate cleavage in acetoclastic methanogenic native to conventional mesophilic anaerobic digestion (35°C). , such as and Methanosaeta, and (ii) non- The majority of systems are operated at 55°C (5, 6), as higher acetoclastic oxidation, i.e., the syntrophic association of acetate- temperatures can result in instability due to ammonia inhibition oxidizing organisms and hydrogenotrophic methanogens, includ- and reduced operability (2). Thermophilic anaerobic digestion ing (5, 16). The latter involves interspecies offers higher rates of destruction of pathogens and organic solids electron transfer, in which acetate oxidizers transfer electrons as well as potentially enhancing methane production. Several through an electron carrier such as hydrogen or formate (17) studies have shown that improvements in performance in ther- mophilic digestion are mainly due to an increase in hydrolysis coefficient (7–9), which determines speed of degradation, rather Received 29 May 2013 Accepted 7 August 2013 than to an increase in the fraction of degradable material. In ad- Published ahead of print 16 August 2013 dition to this advantage, the growth rates of thermophilic meth- Address correspondence to Damien J. Batstone, [email protected]. anogens are 2 to 3 times higher than those of mesophilic meth- Copyright © 2013, American Society for Microbiology. All Rights Reserved. anogens (10). Therefore, the HRT can theoretically be reduced to doi:10.1128/AEM.01730-13 5 to 8 days at 55°C. Recent studies have reported substantial

October 2013 Volume 79 Number 20 Applied and Environmental Microbiology p. 6491–6500 aem.asm.org 6491 Ho et al.

TABLE 1 Characteristics of thickened waste-activated sludge (WAS) hydraulic retention time (HRT) of 4 days throughout the experimental from the Elanora wastewater treatment plant program; the second reactor acted as an experimental system with HRT as Characteristic Valuea the experimental variable. Reactor temperature was controlled by circu- lating hot water inside the reactors’ water jackets from a water bath main- Ϯ Total solids (g/liter) 28.2 0.8 tained at 55°C. Reactors were fed at intervals of4h(6times daily). During Ϯ Volatile solids (g/liter) 19.3 0.5 feeding events, fresh feed sludge was added to the reactors and an equal Ϯ pH 7.1 0.1 volume of effluent was withdrawn simultaneously using a multihead per- Ϯ Chemical oxygen demand (g/liter) 28.7 2.5 istaltic pump. The experimental reactors were operated at three HRTs and Volatile fatty acid (g/liter) 0.2 Ϯ 0.1 ran for a total of 350 days: (i) 4-day HRT (200 days), (ii) 3-day HRT (100 Downloaded from ϩ Ϯ NH4 -N (g/liter N) 0.08 0.03 days), and (iii) 2-day HRT (50 days). a Values are means Ϯ standard deviations for 10 different feed collections used in the Biogas and volumetric methane production rates, pH, volatile fatty study over 12 months. acids (VFAs), and volatile solid (VS) destruction were closely monitored to evaluate process performance. Analytical methods. Digestate was collected from each reactor three while keeping the concentration of the electron carrier low (18). times per week for chemical analyses, including total solids (TS), volatile solids (VS), volatile fatty acid (VFA), chemical oxygen demand (COD), An increase in temperature leads to thermodynamic favorability ϩ and ammonium-nitrogen (NH -N). Analysis followed standard meth- of the oxidation component of the process (19, 20). Hypothesized 4 ods (24). Biogas volume was measured daily from each reactor using http://aem.asm.org/ mechanisms such as direct interspecies electron transfer (17, 21) tipping bucket gas meters and continuously logged online, while gas com- would be similarly enhanced through enhancement of the oxida- position (H2,CO2, and CH4) was determined using a PerkinElmer gas tion component. Next-generation pyrosequencing has further chromatograph (GC) equipped with a thermal conductivity detector clarified methanogenic populations in thermophilic systems (6, (GC-TCD). 15), confirming broad diversity but generally an increased domi- (i) VS destruction. The two calculation methods used to determine VS nance of Methanosarcina over Methanosaeta compared to meso- destruction were the Van Kleeck equation and the mass balance equation. philic systems. The Van Kleeck equation assumes that the amount of mineral solids is Ϫ The relative contribution to methanogenesis via the different conserved during digestion (25) and uses the volatile fractions (VS/TS VSfrac) in the inlet and outlet as references: metabolic pathways can be determined using the stable isotopic on November 12, 2015 by University of Queensland Library 13 12 Ϫ signatures of C/ C. The technique has been used mostly to VSfrac VSfrac ϭ i o ϫ % VS destruction Ϫ ϫ 100 study biogenic methane formation in anoxic environments and VSfrac (VSfrac VSfrac ) has shown that production of ␦CH from reduction of CO exhib- i i o 4 2 where VS and VS are the volatile fractions (VS/TS) in the inlet and fraci fraco its larger fractionation than that from cleavage of acetate (22, 23). outlet solids, respectively. In the context of an anaerobic digester, it can be used to study the The mass balance equation uses VS concentrations in the inlet (feed) relative contribution from acetate cleavage and oxidation, with and outlet (effluent), expressed as 13 ␦13 oxidation resulting in a lower CH4 isotopic signature ( CH4) Ϫ VSconc VSconc due to differential partitioning of bicarbonate to methane during % VS destruction ϭ i o ϫ 100 VS hydrogenotrophic methanogenesis (23). conci where VS and VS are VS concentrations (g/liter) of inlet and In this study, we determined the potential of high-rate, high- conci conco temperature methanogenic digestion as a replacement technology outlet, respectively. for conventional mesophilic digestion of organic solids at the In addition, batch biochemical degradability tests (biochemical meth- most commonly applied thermophilic temperature of 55°C. The ane potential [BMP] tests) were conducted on the inlet and outlet to analysis was based on operation of laboratory-scale continuous provide an independent assessment of process performance, while spe- cific methanogenic activity (SMA) tests were performed to evaluate the anaerobic digesters operating on waste-activated sludge (WAS) specific function of microbial communities within the continuous digest- for approximately 1 year. The HRT was reduced to 2 to 4 days in ers. The test protocols are listed below. order to determine the limitation of methanogenesis. Analytical techniques focused on identifying kinetic capacity, limits, mech- anism, and identity of the functional digester and its community for thermophilic anaerobic digestion.

MATERIALS AND METHODS Substrate. The substrate was biological nutrient removal (BNR) waste- activated sludge (WAS) (10-day aerobic sludge age) collected from the Elanora wastewater treatment plant, located at Gold Coast, Australia. The feed was thickened to obtain total solids (TS) of 2 to 3% and stored at approximately 4°C prior to feeding to minimize degradation and preserve integrity of the material. Seed sludge was collected from a mesophilic anaerobic digester from the Oxley Creek wastewater treatment plant in Brisbane, Australia. Thickened WAS was fed to each single stage reactor in a short pulse every 4 h. The average characteristics of WAS are shown in Table 1. Experimental setup and operational protocol. The experimental platform consisted of two single-stage thermophilic reactors (1-liter working volume) operated in a semicontinuous mode at 55°C and neutral pH (Fig. 1). One reactor served as a control and was operated at a constant FIG 1 Schematic diagram of the bench scale thermophilic reactor.

6492 aem.asm.org Applied and Environmental Microbiology High-Rate Thermophilic Anaerobic Digestion

(ii) Assay for residual biochemical methane potential. Biochemical methane potential (BMP) tests were employed to determine the residual methane potential from digestion residues and assess the stability of the effluent. The method is based on the work of Angelidaki et al. (26). Batch digestions were performed in 160-ml nonstirred glass serum bottles (100-ml working volume) at 37°C. The inoculum was material collected from a mesophilic anaerobic digester operating at a municipal wastewater treatment plant in Brisbane, Australia. Substrates used in the BMP tests were digestates collected from the thermophilic reactors on day 137, day Downloaded from 249, and day 323, respectively. The inoculum to substrate ratio used in the BMP test was 1:1 (VS basis). All tests were carried out in triplicate, and triplicate blanks were conducted to correct for background methane from the inocula. Error bars in the figures represent 95% confidence in error based on the average of the triplicate (two-tailed t test).

The key kinetic parameters of BMP test are degradation extent (fd), which is the fraction of the substrate that may be converted to methane, and the apparent first-order hydrolysis rate coefficient (khyd), which de- http://aem.asm.org/ termines speed of degradation. The parameters fd and khyd were estimated by optimizing fd and khyd in a simple first-order model through Aquasim 2.1d, as was done in the batch tests described by Batstone et al. (27). (iii) Continuous modeling. Continuous reactor modeling was done to determine that measures of degradability sourced through BMP were representative of reactor performance. The test continuous reactor was modeled using the IWA anaerobic digestion model 1 (ADM1) (28)in Aquasim 2.1d, the WAS interface model of Nopens et al. (29), the ob- served COD/VS ratios given in Table 1, and measured biomass nitrogen content. Input material degradability (fd) and hydrolysis coefficient (khyd) on November 12, 2015 by University of Queensland Library were simultaneously estimated using both gas flow and effluent solids concentration (two-parameter searches were done) using secant searches, with parameter uncertainty being reported as linear uncorrelated esti- mates of 95% confidence regions based on two-tailed t tests. Model source files are available on request. FIG 2 Volatile solid destruction for mass balance () and Van Kleeck (Œ) (iv) Specific methanogenic activity (SMA) tests. Methanogenic activ- calculation in the control reactor (top) and test reactor (bottom). VS destruc- ity tests were performed in nonstirred, 160-ml serum bottles incubated at tion in the test reactor when operated at 3- and 4-day HRTs was stable and 55°C. Each bottle contained 10 ml inoculum and 90 ml basic anaerobic comparable to that of the control reactor, but it varied strongly at 2-day HRT (BA) medium (100-ml total liquid volume). BA medium was prepared as (vertical lines indicate transitions from a 4- to a 3- to a 2-day HRT in the test described by Angelidaki et al. (26), while inoculum was harvested from reactor). the continuous thermophilic reactors during the steady state, particularly at day 137 (period I), day 270 (period II), and day 342 (period III). Sodium acetate and sodium formate were added as the substrates to concentra- nogenesis exhibits a significantly larger fractionation factor than that of tions of 3 g/liter and 6 g/liter, respectively, to determine acetoclastic and acetate-dependent methanogenesis due to the ability to select 12C prefer- hydrogenotrophic activity. Blanks contained inoculum and medium entially during methanogenesis (31, 32). The fractionation values mea- without the substrate. Formate was chosen as an alternative source of sured in pure cultures of CO2 reducers (hydrogen utilizers) range between electron donor to H2 in the reduction of CO2, since they are thermody- 1.031 and 1.077 and from 1.007 to 1.027 for the acetoclastic pathway (31). namically and stoichiometrically similar (30). All bottles were flushed ␣ ␦13 ϩ The apparent fractionation ( ) can be calculated as ( CO2 1,000)/ with high-purity N gas for 3 min (1 liter/min), sealed with a rubber ␦13 ϩ 2 ( CH4 1,000). stopper retained with an aluminum crimp cap, and stored in tempera- ␦ The relative contributions of acetate-derived CH4 ( ma) and CO2- ture-controlled incubators at 55°C (Ϯ1°C). All tests were carried out in ␦ ␦ derived CH4 ( mc) to total CH4 ( CH4) can be quantified using the fol- triplicate, and all error bars in the figures indicate 95% confidence in the ␦ ϭ ␦ ϩ Ϫ ␦ lowing mass balance equation: CH4 fmc mc (1 fmc) ma. average of the triplicates. Gas production, composition, and substrate ␦ Ϫ␦ ␦ Ϫ␦ Thus, fmc is solved as follows: ( CH4 ma)/( mc ma). contents were monitored throughout the experimental period. Specific ␦ Ϫ ma was chosen at an average value of 33‰ according to reference methanogenic activity was determined by regression against the linear ␦ 31, whereas mc was chosen based on the dominant acetoclastic methano- section of the methane evolution curve. gens obtained from the pyrotag sequencing results. (v) Stable carbon isotopic signature. The gas samples for stable car- The analyses of isotopic signatures were performed using an Isoprime ␦13 ␦13 bon isotopic analysis of CH4 and CO2 were collected in a time series gas chromatograph combustion isotope ratio mass spectrometer (GC-c- in 5-ml Exetainers during the SMA tests. The principle of 13C/12C frac- IRMS) linked to an Agilent 7890A gas chromatograph with a CP-Poraplot ϫ tionation is discussed by Conrad (31). Isotopic fractionation evaluates Q column (50 m 0.32 mm), with helium as the carrier gas. A CO2-CH4 ␦13 ϭϪ ␦13 ϭ enrichment of one isotope relative to another in specific products and (1:1) gas mixture standard ( CO2 18.76‰ and CH4 infers the generation pathway based on partitioning through the meta- Ϫ40.14‰) was injected before and after gas analysis. Each sample was run 13 bolic process. The fraction of CH4 produced from the CO2 reduction in duplicate, and all reported ␦ C values have a precision with an average ␦ Ϫ␦ ␦ Ϫ␦ pathway to total CH4, fmc, is calculated as ( CH4 ma)/( mc ma), standard deviation of 0.3‰. ␦ ␦ ␦ where CH4 is the measured variable and ma and mc are isotopic values (vi) Molecular analysis. The analysis of microbial community struc- of acetate-derived CH4 and CO2-derived CH4, respectively. ture and diversity were carried out during stable operation using fluores- Generally, microbes preferentially utilize 12C compounds over 13C cence in situ hybridization (FISH) and 16S rRNA gene pyrosequencing. compounds (23). According to the literature, hydrogenotrophic metha- FISH was performed as described by Amann et al. (33). FISH preparations

October 2013 Volume 79 Number 20 aem.asm.org 6493 Ho et al.

TABLE 3 Comparison of degradation parameters of feed WAS and short-HRT digester effluent fitted using model-based analysisa

Material or B0 (liters CH4/kg Ϫ1 HRT khyd (day ) fd VSadded) Feed material 0.19 Ϯ 0.040 0.41 Ϯ 0.18 222 Ϯ 12 (WAS) 4-day HRT 0.18 Ϯ 0.010 0.20 Ϯ 0.011 131 Ϯ 9 3-day HRT 0.19 Ϯ 0.014 0.20 Ϯ 0.010 128 Ϯ 10 Downloaded from 2-day HRT 0.17 Ϯ 0.010 0.31 Ϯ 0.012 170 Ϯ 7 a khyd is the apparent first-order hydrolysis coefficient; fd is the degradation fraction; B0 is the methane production potential. Samples from 4-day, 3-day, and 2-day HRTs were collected on day 137, day 249, and day 323, respectively. Values are means Ϯ standard deviations for triplicates.

tuating VS destruction for the control reactor at an average of 33 http://aem.asm.org/ to 35% using both mass balance and Van Kleeck equations, indi- cating consistency between the two measures. While VS destruc- FIG 3 Gas production in control and experimental reactors. tion showed some variation, this is to be expected where the anal- ysis is momentary. The performance of the test reactor was comparable to that of the control across the first two periods when were visualized with a Zeiss LSM 510 Meta confocal laser scanning micro- HRT was reduced from 4 days to 3 days. However, at a 2-day HRT, scope (CLSM) using three excitation channels (488 nm, green emission; VS destruction varied greatly between the two methods, indicat- 545 nm, red emission; and 633 nm, blue emission). FISH probes used in ing non-steady-state issues.

this study included Eub338 (5=-GCTGCCTCCCGTAGGAGT-3=) on November 12, 2015 by University of Queensland Library mix In most cases, H2 was not detected (and was therefore below probes for most members of the Bacteria (34), Arc915 (5=-GTGCTCC the detection limit of 0.1%), while CH4 accounted for more than CCCG CCAATTCCT-3=) probes for most Archaea (35), Sarci551 (5=-GA 60% of biogas composition and CO for 30%. Total methane pro- CCCAATAATCACGATCAC-3=) for members of the Methanosar- 2 duction levels obtained during the 4-day and 3-day HRT periods cina (36), and Ttoga660 (5=-GTTCCGTCTCCCTCTACC-3=) for members of the order Thermotogales (37). were similar and ranged from 150 to 180 liter CH4/kg VSadded,as Genomic DNA was extracted from the dewatered pellets using a shown in Fig. 3. However, gas production decreased by 30% when FastDNA spin kit for soil. The quantity and quality of the extracted DNA the HRT was shortened to 2 days. The key measure to determine were measured using a NanoDrop ND-1000 spectrophotometer (Nano- process stability was VFA concentration, as shown in Table 2. The Ͻ Drop Technology, Rockland, DE) and agarose gel (0.8%, wt/vol) electro- concentrations of C2-C4 VFAs were low ( 200 mg/liter COD) at a phoresis. The primers used for pyrotag sequencing were 926f (5=-AAAC 4-day HRT in both the control and test reactors and then in- TYAAAKGAATTGACGG-3=) and 1392r (5=-ACGGGCGGTGTGTAC- creased consistently to approximately 500 mg/liter COD at 3=)(33). 16S rRNA gene pyrosequencing was carried out according to shorter HRTs (Table 2). pH values of the digesters decreased from Roche 454 protocols using a Roche 454 GS FLX sequencer (Roche, Swit- 7.04 during the 4-day HRT to 6.89 during the 3-day HRT and zerland). Sequences that were shorter than 350 bp and reads containing varied in the range of 6.74 to 7.04 during the 2-day HRT. The any unresolved nucleotides were removed from the pyrosequencing-de- rived data sets. Operational taxonomic units (OTUs) were determined for ammonia concentrations in both digesters varied in the range of each denoised sequence data set by using the uclust OTU picker version 700 to 1,000 mg/liter and were not affected by changes in HRT. 1.2.21q of the QIIME software pipeline (38). Digestate degradability. The digestates produced at different HRTs were assessed using BMP tests. A summary of the hydrolysis RESULTS rate coefficients and degradability fractions determined using the Reactor performance. The process performance of both control nonlinear parameter estimation method is shown in Table 3, and experimental reactors was evaluated by VS destruction, pro- while the cumulative methane production is shown in Fig. 4. The duction of methane, and organic acid levels. Figure 2 shows fluc- degradability of feed sludge was estimated at approximately 41%,

TABLE 2 Comparison of levels of volatile fatty acids between the control and test reactors across three operating periods Value (mg/liter COD)a

Period HRT Total VFAs Acetate Propionate C4ϩ acids I Control 175 Ϯ 68 97 Ϯ 28 65 Ϯ 24 13 Ϯ 7 4 day 104 Ϯ 53 67 Ϯ 30 32 Ϯ 23 5 Ϯ 2

II Control 164 Ϯ 43 104 Ϯ 18 51 Ϯ 28 10 Ϯ 3 3 day 254 Ϯ 82 171 Ϯ 49 77 Ϯ 30 5 Ϯ 3

III Control 98 Ϯ 19 79 Ϯ 17 11 Ϯ 48Ϯ 2 2 day 497 Ϯ 224 301 Ϯ 182 142 Ϯ 67 54 Ϯ 28 a Values are means Ϯ standard deviations for different measurements over each period.

6494 aem.asm.org Applied and Environmental Microbiology High-Rate Thermophilic Anaerobic Digestion

TABLE 4 Summary of kinetic parameters for the acetate-utilizing a (km,ac) and hydrogenotrophic (km,for) methanogenesis Value (g COD · g VSϪ1 · dayϪ1)

HRT km,ac km,for 4 day 0.77 Ϯ 0.022 1.4 Ϯ 0.13 3 day 0.69 Ϯ 0.018 1.5 Ϯ 0.17 2 day 0.81 Ϯ 0.069 1.4 Ϯ 0.20

a Samples from 4-day, 3-day, and 2-day HRTs were collected on day 137, day 270, and Downloaded from day 342, respectively. Values are means Ϯ standard deviation for triplicates.

selves made an additional 10% available for degradation, as seen by the drop in the recalcitrant component. Model-based analysis results in Fig. 5 show that the BMP was a good estimate of in-

reactor performance to determine fd, as the input degradability estimated based on model outputs aligned well with material as http://aem.asm.org/

analyzed by BMP. Uncertainty in model-based estimates of fd was high due to correlation with khyd and the limited amount of dy- FIG 4 Cumulative methane production from the residual degradation tests. namic operation. k was estimated at 1.5 and 2.5 dayϪ1 (solids Points represent experimental measurements, error bars show 95% confi- hyd dence intervals based on triplicate analyses, and lines represent model fit. and gas, respectively), though with an uncertainty on the order of Ϯ1 dayϪ1. Specific methanogenic activity (SMA) tests. The acetate up- take rate was in the range of 0.69 to 0.81 g COD·gVSϪ1 · dayϪ1, with an average gas production of 210 to 230 liter CH4/kg VSadded Ϫ1 while formate uptake was double that, at 1.4 to 1.5 g COD·gVS · on November 12, 2015 by University of Queensland Library over a 40-day incubation. The values were in agreement with pre- dayϪ1 (Table 4). There were no significant differences in the substrate vious studies (4, 7). uptake rates between digesters with different retention times. Blanks Figure 5 shows an analysis of the residual methane potential of are also shown (Fig. 6), demonstrating that endogenous methane the digestates collected from each reactor compared to that of feed production was negligible compared with substrate-derived methane sludge. Recalcitrant component is based on the material not con- Ϫ production. verted to methane (1 fd). The material calculated as destroyed in The stable carbon isotopic signatures of CH4/CO2 and the frac- reactors is the difference between feed and digestate potentials. tionation factors of acetate and formate degradation are presented This showed that the 4-day and 3-day HRT continuous thermo- in Fig. 6. The results indicate that formate was degraded, and as philic reactors degraded 80% of the available material, while the ␦ ␦ 12 Ϫ expected, CO2 and CH4 were both enriched in C( 18‰ and 2-day HRT system converted only 50% of the available material, Ϫ60‰, respectively). The isotopic signatures became more neg- leaving a less stabilized product. The thermophilic reactors them- ative (i.e., further enriched in 12C) with increasing uptake of for- mate. In contrast, the produced methane become increasingly heavier during acetate conversion, suggesting a gradual shift of the dominant pathway from hydrogenotrophic to acetoclastic metha- nogenesis as the reaction proceeded. In all cases, except at the end of the acetate batch (last two points), the majority of the methane was apparently hydrogen derived. Community structure. FISH analyses were carried out to de- termine the active methanogenic and bacterial populations. Flu- orescence microscopy of FISH samples collected at different time points always found that the archaea were dominated by Metha- nosarcina (Fig. 7, pink) with a complete overlap between total archaea (Arc915-cy3 probe; Fig. 7, red) and Methanosarcina-spe- cific (Sarci551-cy5 probe; Fig. 7, blue). However, during the 2-day HRT period, the portion of Methanosarcina was much lower, sug- gesting a decrease in its relative abundance. Instead, we found thermophilic bacteria in the order Thermotogales (Fig. 7, yellow,

representing the overlap between the Eub338mix-FITC [fluores- FIG 5 Comparison of degradable fractions of feed WAS and short HRT di- cein isothiocyanate] and Ttoga660-cy3 probes). gesters (error bars show 95% confidence in mean VS destruction, based on Pyrotag sequencing results confirmed the dominance of mem- triplicate analyses of the BMP tests). Potentials for 4, 3, and 2 days were taken bers of the archaeal family Methanosarcinaceae, with members of from the experimental reactor on day 137, day 249, and day 323, respectively. the Methanobacteriaceae also being present at approximately 15 to Estimates of in-reactor performance were based on averages Ϯ errors of two data sets, i.e., gas data and VS data, for 6 measurements around these days. 20% of the methanogenic population (Fig. 8). The bacterial community was composed mainly of the phyla Proteobacteria (no- Continuous results based on parameter estimation (khyd and fd) with simula- tion of whole data set. tably Alphaproteobacteria and Betaproteobacteria), Bacteroidetes

October 2013 Volume 79 Number 20 aem.asm.org 6495 Ho et al. Downloaded from http://aem.asm.org/ on November 12, 2015 by University of Queensland Library

␦13 FIG 6 Evolution of CofCH4 and CO2 and the fraction of CH4 produced from CO2 reduction pathway (fmc) during acetate (top) and formate (bottom) conversion (note that the time scales for acetate and formate are different). Error bars show 95% confidence based on triplicate analyses; solid lines represent ␦ ␦ CO2 and CH4; dashed lines represent fmc. SMA based on samples taken on day 137 (4-day HRT).

(mainly members of the class Sphingobacteriia), and Chloroflexi DISCUSSION (dominated by members of the class Anaerolineae), which were The overall performance of the high-rate thermophilic anaerobic abundant in the feed sludge. There was a significant increase in digesters in this study was comparable to the results obtained from members of the phylum Firmicutes (20 to 30%), affiliated at the mesophilic digesters operated at 15 to 20 day HRT reported by Ge order level with Clostridiales, in the digester samples and a notice- et al. (7) and Bolzonella et al. (39) and better than that of digesters able presence of members of the order Thermotogales (5 to 10%) in operated at reduced HRTs (8 to 12 days) reported by Lee et al. (12) the microbial communities harvested from the digesters at shorter and Nges and Liu (40). Contradictory to these previous studies, HRTs. This is consistent with FISH analyses and previous findings methanogenesis of the high-rate system in the current study re- (6, 20). mained relatively stable, with a consistently low VFA content of The methanogenic population dropped from 25 to 30% to less Ͻ200 mg/liter at 3- and 4-day HRTs. In previous studies, a short than 5% of the total microbial population as the HRT was short- HRT of 2 to 4 days was used mainly in the thermophilic pretreat- ened from 4 to 2 days. The bacterial community was mostly influ- ment stage to intensify hydrolysis rather than methanogenesis; enced by the feed microbial composition, but the level of diversity thus, gas production was relatively low (11, 39). This is not the varied in response to operating conditions, in this case HRT. The case for our system, and reasons for this may include the consis- microbial communities obtained at 4-day (day 25 to 200) and tent and stable start-up and operation at a low HRT, which se- 3-day (day 245 to 287) HRTs were similar and remained consis- lected from the start for competent organisms in this context. tent throughout the operating periods. During the 2-day HRT More than 80% of the available material was degraded and subse- period (day 300 to 350), the bacterial population changed strongly quently converted to methane at the HRT of 3 to 4 days, which is over time, which likely contributed to the large variation in VS consistent with a high hydrolysis coefficient. destruction and process instability under these conditions. The The stable methanogenesis achieved in this study can be attrib- key functional groups were still prevalent, with a noticeable shift uted to higher reaction rates at elevated temperatures and changes in in abundance toward thermophilic members of the Firmicutes and microbial community structure and their metabolic pathways in re- Thermotogae and a drop in the phylum Chloroflexi. sponse to thermodynamic changes and short-retention-time pres-

6496 aem.asm.org Applied and Environmental Microbiology High-Rate Thermophilic Anaerobic Digestion Downloaded from http://aem.asm.org/ on November 12, 2015 by University of Queensland Library

FIG 7 FISH images of the microbial consortia derived from waste-activated sludge (a), the thermophilic digester at day 200 (4-day HRT) (b), the thermophilic digester at day 285 (3-day HRT) (c), and the digester at day 330 (2-day HRT) (d). Bacteria (detected with the probe Eub338mix) are in green, aggregates of Methanosarcina (Sarci551) are in pink, and rod-shaped Thermotogales (Ttoga660) are in yellow.

sure. As reported earlier (3, 7), the hydrolysis rate was almost double shift from CO2-reducing methanogenesis to acetate cleavage at when the temperature was increased from 35°C to 55°C, and the the end of the batch test. However, the continuous system was improvement was transferred to the subsequent processes. Likewise, consistently dominated by oxidation, with fmc being larger than apparent methanogenic coefficients increased from 0.37 dayϪ1 under the theoretical 33% (41), possibly due to constant washout and mesophilic conditions (3) to 0.77 dayϪ1 for acetoclastic - replacement of the microbial communities in the continuous di- esis and 1.4 dayϪ1 for hydrogenotrophic methanogenesis under ther- gesters. This observation suggests that cleavage may dominate mophilic digestion. Uptake on formate (as a proxy for hydro- again at longer retention times. The thermodynamic shifts in genotrophic methanogenesis) was higher than on acetate, which is in methanogenic pathways with respect to increase in temperature agreement with previous studies (3, 14). The results therefore indi- have also been reported (16, 42), suggesting that high temperature cated that hydrolysis and methanogenesis were both improved at and short HRT favor the syntrophic association between acetate 55°C. With a 2-day HRT, methanogenic coefficients were similar to oxidizers and hydrogenotrophic methanogens, which may in- the values obtained with 3- and 4-day HRTs, yet the overall perfor- clude Methanosarcinaceae and Methanobacteriaceae. Methanobac- mance of the continuous digesters decreased. The main failure mode teriaceae have previously been found only to be hydrogen and with a low HRT was loss of methanogenesis, rather than loss of the formate utilizing, while Methanosarcinaceae can also cleave ace- primary hydrolysis, evidenced by a drop in gas flow and accumula- tate, with the metabolic possibility also for acetate oxidation to tion of acetate and C3ϩ acids. This indicates that methanogenesis is hydrogen (43, 44). The high level of formate activity may be due to the rate-limiting step at high temperatures (rather than hydrolysis) activity by Methanobacteriaceae, since Methanosarcinaceae have and that unlike in many mesophilic systems, hydraulic failure can be not been previously observed to grow on formate (though the determined by an increase in intermediates rather than unhydrolyzed genes for this do exist [45]). We believe part of the reason the material simply washing out. reactors were stable at low HRT is that applying a short HRT from When examining the methanogenic pathways in particular us- the start enabled a stable dominance of oxidation capacity, which ing isotopic fractionation analysis (4 day HRT), we observed a was further favored by the higher temperature.

October 2013 Volume 79 Number 20 aem.asm.org 6497 Ho et al. Downloaded from http://aem.asm.org/ on November 12, 2015 by University of Queensland Library

FIG 8 Relative abundances of phylogenetic groups in the thermophilic digesters collected at different operating times and then grouped in terms of HRTs. Bacterial communities are presented by the dominating phyla, while the archaeal are presented by the dominant families Methanosarcinaceae and Methanobacteriaceae. Phylogenetic groups accounting for Յ0.5% of all classified sequences are summarized in the artificial group “others.”

The microbial community structures of mesophilic anaerobic of methanogens rather than the reduced activity of the syntrophs. digesters have been well documented with the dominant bacterial Indeed, if syntrophic bacterial acetate oxidizers were lost while groups of Chloroflexi, Bacteroidetes, and Firmicutes and archaeal Methanosarcinaceae were retained, one would expect that methano- members of and Methanosaetaceae (13, 46). genesis would be maintained through acetoclastic methanogenesis However, the pyrotag sequencing results showed the exclusive rather than acetate oxidation, which was not the case. dominance of Methanosarcinaceae and the hydrogenotrophic A significant reduction of HRT from 15 to 20 days down to 3 to methanogens Methanomicrobiaceae, including the genera of 4 days means that the daily sludge flow rate could be increased up Methanobrevibacter, Methanothermobacter, and Methanobacte- to five times or the digester volume could be likewise reduced. The rium, in the thermophilic digesters. Methanosarcinaceae, which heating demand for thermophilic digesters operated at 55°C can dominate in the thermophilic digester, are capable of either hy- be provided from methane production using cogeneration. Fur- drogenotrophic or acetoclastic methanogenesis. The absence of thermore, where semibatch operation is applied, the material obligate acetoclastic Methanosaetaceae could be attributed to a meets the standards for pathogen destruction (51). The VS de- high ammonia level (5) and short HRT, which imposes selective struction rate of the thermophilic digester was 33 to 35%, which is pressure on the slow-growing microorganisms (the maximum comparable to what would be achieved at approximately 12 to 15 Ϫ Ϫ growth rates were 0.2 day 1 for Methanosaetaceae and 0.6 day 1 days in a mesophilic digester (based on BMP results). Residual for Methanosarcinaceae [3, 47]). degradability at 4 days and 3 days was 0.20, which is comparable to Only a limited number of syntrophic acetate oxidation bacteria the commonly applied standard of 17% (52) and indicates that the have been isolated and characterized, including the thermophilic material is suitable for reuse either immediately or after additional bacteria Thermacetogenium phaeum (48), belonging to the phy- minor conditioning. lum Firmicutes, Thermotoga lettingae strain TMO (49), belonging to the phylum Thermotogae, and two mesophilic bacteria both ACKNOWLEDGMENTS belonging to the phylum Firmicutes (50). These phyla were dom- inant in the thermophilic biogas reactors, with increasing abun- Dang P. Ho is an Environmental Biotechnology Cooperative Research dance of the orders Clostridiales and Thermotogales at shorter Centre Ph.D. student through project P23, and the project received fund- ing support from the EBCRC. Damien J. Batstone is the recipient of an HRTs. While hypothetical acetate-oxidizing members of the Chlo- ARC Research Fellowship, and this project was supported under Austra- roflexi were lost at a 2-day HRT, this appears to have been offset by lian Research Council’s Discovery Projects funding scheme (project num- the increase in other acetate-oxidizing candidates. In fact, the ber DP0985000). Paul D. Jensen is an MLA/AMPC research fellow. drop in Methanosarcinaceae (from almost 20% to Ͻ5%) was the We thank Beatrice-Keller Lehmann and Kim Baublys from the Uni- most distinguishable microbial characteristic of the change from pre- versity of Queensland for assistance with analysis. to postfailure. It suggests that the system collapsed mainly due to loss We have no competing financial interests to declare.

6498 aem.asm.org Applied and Environmental Microbiology High-Rate Thermophilic Anaerobic Digestion

REFERENCES 23. Laukenmann S, Polag D, Heuwinkel H, Creule M, Gronauer A, Le- 1. Tchobanoglous G, Burton FL, Stensel HD. 2003. Wastewater engineer- lieveld J, Keppler F. 2010. Identification of methanogenic pathways in ing: treatment and reuse, 4th ed. McGraw-Hill, New York, NY. anaerobic digesters using stable carbon isotopes. Eng. Life Sci. 10:509– 2. Batstone DJ, Jensen PD. 2011. Anaerobic processes, p 615–639. In Wil- 514. derer P, ed. Treatise on water science. Elsevier B.V., Amsterdam, The 24. APHA. 2005. Standard methods for the examination of water & wastewa- Netherlands. ter, 21st ed. American Public Health Association, Washington, DC. 3. Siegrist H, Vogt D, Garcia-Heras JL, Gujer W. 2002. Mathematical 25. Switzenbaum MS, Farrell JB, Pincince AB. 2003. Relationship between the Van Kleeck and mass-balance calculation of volatile solids loss. Water model for meso- and thermophilic anaerobic sewage sludge digestion. Environ. Res. 75:377–380. Environ. Sci. Technol. 36:1113–1123. 26. Angelidaki I, Alves M, Bolzonella D, Borzacconi J, Campose L, Guwy Downloaded from 4. Speece RE. 2008. Anaerobic biotechnology and odor/corrosion control AJ, Kalyuznyi S, Jenicek P, Van Lier JB. 2009. Defining the biomethane for municipalities and industries. Archae Press, Nashville, TN. potential (BMP) of solid organic wastes and energy crops: a proposed 5. Karakashev D, Batstone DJ, Trably E, Angelidaki I. 2006. Acetate protocol for batch assays. Water Sci. Technol. 59:927–934. oxidation is the dominant methanogenic pathway from acetate in the 27. Batstone DJ, Tait S, Starrenburg D. 2009. Estimation of hydrolysis param- absence of Methanosaetaceae. Appl. Environ. Microbiol. 72:5138–5141. eters in full-scale anerobic digesters. Biotechnol. Bioeng. 102:1513–1520. 6. Sundberg C, Al-Soud WA, Larsson M, Alm E, Yekta SS, Svensson BH, 28. Batstone DJ, Keller J, Angelidaki I, Kalyuznyi S, Pavlostathis SG, Rozzi Sorensen SJ, Karlsson A. 2013. 454 pyrosequencing analyses of bacterial A, Sanders WTM, Siegrist H, Vavilin VA. 2002. IWA Task Group for and archaeal richness in 21 full-scale biogas digesters. FEMS Microbiol. Mathematical Modelling of Anaerobic Digestion Processes. IWA Publish- Ecol. 85:612–626. ing, London, United . 7. Ge H, Jensen PD, Batstone DJ. 2011. Temperature phased anaerobic 29. Nopens I, Batstone DJ, Copp JB, Jeppsson U, Volcke E, Alex J, Van- http://aem.asm.org/ digestion increases apparent hydrolysis rate for waste activated sludge. rolleghem PA. 2009. An ASM/ADM model interface for dynamic plant- Water Res. 45:1597–1606. wide simulation. Water Res. 43:1913–1923. 8. Song Y-C, Kwon S-J, Woo J-H. 2004. Mesophilic and thermophilic 30. Batstone DJ, Keller J, Newell RB, Newland M. 2000. Modelling anaer- temperature co-phase anaerobic digestion compared with single-stage obic degradation of complex wastewater. II: parameter estimation and mesophilic- and thermophilic digestion of sewage sludge. Water Res. 38: validation using slaughterhouse effluent. Biores. Technol. 75:75–85. 1653–1662. 31. Conrad R. 2005. Quantification of methanogenic pathways using stable 9. Kim JK, Oh BR, Chun YN, Kim SW. 2006. Effects of temperature and carbon isotopic signatures: a review and a proposal. Org. Geochem. 36: hydraulic retention time on anaerobic digestion of food waste. J. Biosci. 739–752. Bioeng. 102:328–332. 32. Whiticar MJ, Faber E, Schoell M. 1986. Biogenic methane formation in 10. Van Lier JB, Hulsbeek J, Stams AJ, Lettinga G. 1993. Temperature marine and freshwater environments: CO reduction vs. acetate fermen-

2 on November 12, 2015 by University of Queensland Library susceptibility of thermophilic methanogenic sludge: implications for re- tation-isotope evidence. Geochim. Cosmochim. Acta 50:693–709. actor start-up and operation. Biores. Technol. 43:227–235. 33. Amann RI, Ludwig W, Schleifer KH. 1995. Phylogenetic identification 11. Ge H, Jensen PD, Batstone DJ. 2011. Increased temperature in the and in situ detection of individual microbial cells without cultivation. thermophilic stage in temperature phased anaerobic digestion (TPAD) Microbiol. Rev. 59:143–169. improves degradability of waste activated sludge. J. Hazard. Mater. 187: 34. Stahl DA, Amann RI. 1991. Development and application of nucleic acid 355–361. probes, p 205–248. In Stackebrandt E, Goodfellow M (ed), Nucleic acid 12. Lee I-S, Parameswaran P, Rittmann BE. 2011. Effects of solids retention techniques in bacterial systematics. John Wiley & Sons Ltd., Chichester, time on methanogenesis in anaerobic digestion of thickened mixed England. sludge. Biores. Technol. 102:10266–10272. 35. Raskin L, Stromley JM, Rittmann BE, Stahl DA. 1994. Group-specific 13. Riviere D, Desvignes V, Pelletier E, Chaussonnerie S, Guermazi S, 16S rRNA hybridization probes to describe natural communities of meth- Weissenbach J, Li T, Camacho P, Sghir A. 2009. Towards the definition anogens. Appl. Environ. Microbiol. 60:1232–1240. of a core of microorganisms involved in anaerobic digestion of sludge. 36. Sorensen A, Torsvik V, Torsvik T, Poulsen L, Ahring B. 1997. Whole- ISME J. 3:700–714. cell hybridization of Methanosarcina cells with two new oligonucleotide 14. Regueiro L, Veiga P, Figueroa M, Alonso-Gutierrez J, Stams AJM, Lema probes. Appl. Environ. Microbiol. 63:3043–3050. JM, Carballa M. 2012. Relationship between microbial activity and mi- 37. Harmsen H, Prieur D, Jeanthon C. 1997. Group-specific 16S rRNA- crobial community structure in six full-scale anaerobic digesters. Micro- targeted oligonucleotide probes to identify thermophilic bacteria in ma- biol. Res. 167:581–589. rine hydrothermal vents. Appl. Environ. Microbiol. 63:4061–4068. 15. Demirel B, Scherer P. 2008. The roles of acetotrophic and hydro- 38. Caporaso JG, Kuczynski J, Stombaugh J, Bittinger K, Bushman FD, genotrophic methanogens during anaerobic conversion of biomass to Costello EK, Fierer N, Pena AG, Goodrich JK, Gordon JI, Huttley GA, methane: a review. Rev. Environ. Sci. Biotechnol. (N.Y.) 7:173–190. Kelley ST, Knights D, Koenig JE, Ley RE, Lozupone CA, McDonald D, 16. Sasaki D, Hori T, Haruta S, Ueno Y, Ishii M, Igarashi Y. 2011. Metha- Muegge BD, Pirrung M, Reeder J, Sevinsky JR, Turnbaugh PJ, Walters nogenic pathway and community structure in a thermophilic anaerobic WA, Widmann J, Yatsunenko T, Zaneveld J, Knight R. 2010. QIIME digestion process of organic solid waste. J. Biosci. Bioeng. 111:41–46. allows analysis of high-throughput community sequencing data. Nat. 17. Morita M, Malvankar NS, Franks AE, Summers ZM, Giloteaux L, Methods 7:335–336. Rotaru AE, Rotaru C, Lovley DR. 2011. Potential for direct interspecies 39. Bolzonella D, Pavan P, Battistoni P, Cecchi F. 2005. Mesophilic anaerobic electron transfer in methanogenic wastewater digester aggregates. mBio digestion of waste activated sludge: influence of the solid retention time in the 2:e00159–11. doi:10.1128/mBio.00159-11. wastewater treatment process. Process Biochem. 40:1453–1460. 18. Lee MJ, Zinder SH. 1988. Hydrogen partial pressures in a thermophilic 40. Nges IA, Liu J. 2010. Effects of solid retention time on anaerobic digestion acetate-oxidizing methanogenic coculture. Appl. Environ. Microbiol. 54: of dewatered-sewage sludge in mesophilic and thermophilic conditions. 1457–1461. Renewable Energy 35:2200–2206. 19. Petersen SP, Ahring BK. 1991. Acetate oxidation in a thermophilic an- 41. Penning H, Conrad R. 2007. Quantification of carbon flow from stable aerobic sewage-sludge digestor: the importance of non-aceticlastic metha- isotope fractionation in rice field soils with different organic matter con- nogenesis from acetate. FEMS Microbiol. Lett. 86:149–152. tent. Org. Geochem. 38:2058–2069. 20. Goberna M, Insam H, Franke-Whittle IH. 2009. Effect of biowaste 42. Hao L-P, Lui F, He P-J, Li L, Shao L-M. 2011. Predominant contribution sludge maturation on the diversity of thermophilic bacteria and archaea in of syntrophic acetate oxidation to thermophilic methane formation at an anaerobic reactor. Appl. Environ. Microbiol. 75:2566–2572. high acetate concentrations. Environ. Sci. Technol. 45:508–513. 21. Rotaru A-E, Shrestha PM, Liu F, Ueki T, Nevin K, Summers ZM, Lovley 43. Valentine DL, Blanton DC, Reeburgh WS. 2000. Hydrogen production DR. 2012. Interspecies electron transfer via hydrogen and formate rather by methanogens under low-hydrogen conditions. Arch. Microbiol. 174: than direct electrical connections in cocultures of Pelobacter carbinolicus 415–421. and Geobacter sulfurreducens. Appl. Environ. Microbiol. 78:7645–7651. 44. Heimann AC, Batstone DJ, Jakobsen R. 2006. Methanosarcina spp. drive 22. Fey A, Claus P, Conrad R. 2004. Temporal change of 13C-isotope sig- vinyl chloride dechlorination via interspecies hydrogen transfer. Appl. natures and methanogenic pathways in rice field soil incubated anoxically Environ. Microbiol. 72:2942–2949. at different temperatures. Geochim. Cosmochim. Acta 68:293–306. 45. Maeder DL, Anderson I, Brettin TS, Bruce DC, Gilna P, Han CS,

October 2013 Volume 79 Number 20 aem.asm.org 6499 Ho et al.

Lapidus A, Metcalf WW, Saunders E, Tapia R, Sowers KR. 2006. The 49. Balk M, Weijma J, Stams AJ. 2002. Thermotoga lettingae sp. nov., a novel genome: comparative analysis with Methanosar- thermophilic, methanol-degrading bacterium isolated from a thermo- cina acetivorans and Methanosarcina mazei reveals extensive rearrange- philic anaerobic reactor. Int. J. Syst. Evol. Microbiol. 52:1361–1368. ment within methanosarcinal genomes. J. Bacteriol. 188:7922–7931. 50. Schnürer A, Zellner G, Svensson BH. 1999. Mesophilic syntrophic ace- 46. Westerholm M, Roos S, Schnurer A. 2010. Syntrophaceticus schinkii gen. tate oxidation during methane formation in biogas reactors. FEMS Mi- nov., sp. nov., an anaerobic, syntrophic acetate-oxidizing bacterium iso- crobiol. Ecol. 29:249–261. lated from a mesophilic anaerobic filter. FEMS Microbiol. Lett. 309:100– 51. Paul E, Carrere H, Batstone DJ. Thermal methods to enhance biological 104. treatment processes, p 373–404. In Paul E, Liu Y (ed), Biological sludge 47. Vavilin VA, Qu X, Mazeas L, Lemunier M, Duquennoi C, He P, minimisation and biomaterials/bioenergy recovery technologies. John

Bouchez T. 2008. Methanosarcina as the dominant aceticlastic methano- Wiley & Sons, Inc., New York, NY. Downloaded from gens during mesophilic anaerobic digestion of putrescible waste. Antonie 52. US Environmental Protection Agency. 1994. Pathogen and vector attrac- Van Leeuwenhoek 94:593–605. tion reduction requirements, p 107–127. In A plain English guide to the 48. Hattori S. 2008. Syntrophic acetate-oxidizing microbes in methanogenic EPA part 503 biosolids rule. EPA document 8322/R-93-003. US Environ- environments. Microbes Environ. 23:118–127. mental Protection Agency, Washington, DC. http://aem.asm.org/ on November 12, 2015 by University of Queensland Library

6500 aem.asm.org Applied and Environmental Microbiology